ZMIME
Comparison · 3 models · Updated Nov 18, 2026

Nova Pro vs Claude Haiku 3 vs Qwen2.5-VL 72B Instruct

No independent capability benchmark covers every model here yet. The score uses price, inputs & features and context window.

Price in rating: Included · Exclude price

  1. Amazon

    Nova Pro

    Released Dec 3, 2024

    51/100
    • ECI—
    • Price$0.80 / $3.20
    • Context300K
  2. Anthropic

    Claude Haiku 3

    Released Mar 13, 2024

    57/100
    • ECI117.8
    • Price$0.25 / $1.25
    • Context200K
  3. Alibaba (Qwen)

    Qwen2.5-VL 72B Instruct

    Released Jan 26, 2025

    45/100
    • ECI—
    • Price$1.01 / $1.01
    • Context131K
01 — Verdict

Capability evidence is incomplete

There is not enough shared capability evidence to name an overall winner. The scores (Claude Haiku 3 57/100, Nova Pro 51/100, Qwen2.5-VL 72B Instruct 45/100) compare the available specifications and price. The score weighs price 50%, inputs & features 30%, context window 20%. None of the shared benchmarks cover every model here yet, so capability is left out of this verdict.

  • CapabilityNot enough dataNo independent benchmark covers every model here yet
  • Lowest priceClaude Haiku 3Claude Haiku 3 $0.50 · Qwen2.5-VL 72B Instruct $1.01 · Nova Pro $1.40 per 1M tokens (3:1 blend)
  • Longest contextNova ProNova Pro 300,000 · Claude Haiku 3 200,000 · Qwen2.5-VL 72B Instruct 131,072 tokens
  • Widest inputsNova ProNova Pro: Text, Images, PDFs, Video · Claude Haiku 3: Text, Images, PDFs · Qwen2.5-VL 72B Instruct: Text, Images
  • Self-hostingQwen2.5-VL 72B InstructPublishes downloadable weights
How the score is built
MeasureWeightNova ProClaude Haiku 3Qwen2.5-VL 72B Instruct
Price50%416148
Inputs & features30%706050
Context window20%453932
Specs & price only100%51/10057/10045/100

Left out because at least one model lacks the data: capability. The remaining weights were rescaled.

02 — Side by side

Every spec in one table

Highlighted cells lead their row. Dashes mean the data is not published.

Nova Pro vs Claude Haiku 3 vs Qwen2.5-VL 72B Instruct specifications side by side
SpecificationNova ProAmazonClaude Haiku 3AnthropicQwen2.5-VL 72B InstructAlibaba (Qwen)
Capability
Capabilities Index (ECI)—117.8—
ECI rank—#154 of 158—
GPQA DiamondGraduate-level science questions—36.3%—
OTIS Mock AIME 2024–2025Competition mathematics—1.8%—
Price per million tokens
Input$0.80$0.25 (best)$1.01
Output$3.20$1.25$1.01 (best)
Cached input$0.20——
Blended (3:1)$1.40$0.50 (best)$1.01
Long-context rateSame rateSame rateSame rate
Price sourceOfficial Amazon Bedrock APIMedian of 3 providersMedian of 5 providers
Limits
Context window300,000 tokens (best)200,000 tokens131,072 tokens
Max output10,000 tokens (best)4,096 tokens8,192 tokens
Inputs and features
TextYesYesYes
ImagesYesYesYes
PDFsYesYesNo
AudioNoNoNo
VideoYesNoNo
ReasoningNoNoNo
Tool callingYesYesYes
Structured outputNoNoNo
Availability
WeightsProprietaryProprietaryOpen
API model IDamazon.nova-pro-v1:0——
API providers9 (best)35
ReleasedDec 3, 2024Mar 13, 2024Jan 26, 2025
Knowledge cutoffOct 2024Aug 31, 2023Apr 2024
03 — Cost

What would a month cost?

Enter your expected volume in millions of tokens. List prices only; caching and batch discounts would lower these.

  • Nova Pro$14.40
  • Claude Haiku 3$5.00
  • Qwen2.5-VL 72B Instruct$12.12
04 — Questions

Which should you choose?

Which is better: Nova Pro, Claude Haiku 3 or Qwen2.5-VL 72B Instruct?

There is not enough shared capability evidence to name an overall winner. The scores (Claude Haiku 3 57/100, Nova Pro 51/100, Qwen2.5-VL 72B Instruct 45/100) compare the available specifications and price. The score weighs price 50%, inputs & features 30%, context window 20%. None of the shared benchmarks cover every model here yet, so capability is left out of this verdict.

Which is cheaper, Nova Pro, Claude Haiku 3 or Qwen2.5-VL 72B Instruct?

Claude Haiku 3 is cheaper at $0.25 input / $1.25 output per million tokens (median across 3 API providers). Qwen2.5-VL 72B Instruct costs $1.01 input / $1.01 output per million tokens (median across 5 API providers); Nova Pro costs $0.80 input / $3.20 output per million tokens (official Amazon Bedrock API price). At a typical mix of three input tokens to one output token, that is $0.50 per million tokens for Claude Haiku 3 versus $1.01 for Qwen2.5-VL 72B Instruct (2× as much) and $1.40 for Nova Pro (2.8× as much).

Which scores higher on benchmarks?

There is no independent benchmark that covers all three models yet. Nova Pro has not been scored yet, Claude Haiku 3 has an ECI of 117.8 and Qwen2.5-VL 72B Instruct has not been scored yet.

Which is better for coding?

There are no published SWE-bench Verified results for Nova Pro, Claude Haiku 3 and Qwen2.5-VL 72B Instruct yet, so there is no like-for-like coding score. Test both on a sample of your own tasks. All three support tool calling for agent workflows.

Which has the bigger context window?

Nova Pro has the largest context window at 300,000 tokens, against 200,000 for Claude Haiku 3 and 131,072 for Qwen2.5-VL 72B Instruct. Maximum output per response: Nova Pro up to 10,000, Claude Haiku 3 up to 4,096, Qwen2.5-VL 72B Instruct up to 8,192 tokens.

Which can read images, PDFs, audio or video?

Nova Pro accepts text, images, PDFs and video; Claude Haiku 3 accepts text, images and PDFs; Qwen2.5-VL 72B Instruct accepts text and images. Nova Pro handles the widest range of inputs.

Are any of these open source?

Qwen2.5-VL 72B Instruct publishes its weights and can be self-hosted; Nova Pro and Claude Haiku 3 is proprietary.

Which is newer?

Qwen2.5-VL 72B Instruct is the newest, released Jan 26, 2025. Nova Pro came out Dec 3, 2024; Claude Haiku 3 came out Mar 13, 2024. Knowledge cutoff: Nova Pro Oct 2024, Claude Haiku 3 Aug 31, 2023, Qwen2.5-VL 72B Instruct Apr 2024.

How do you decide the winner?

This rating uses price (50%, blended price per million tokens (3 input : 1 output), log scale); inputs & features (30%, image, PDF, audio and video input, tool calling, structured output and reasoning); context window (20%, maximum tokens per request, log scale). Dimensions missing for any model are dropped and the remaining weights rescaled, so every model is judged on the same evidence. Specs and prices come from public model listings and the labs’ own API pages; capability scores come from independent benchmark runs. Data updated Nov 18, 2026.